European Journal of Public Health, Vol. 16, No. 6, 670–675 Ó The Author 2006. Published by Oxford University Press on behalf of the European Public Health Association. All rights reserved. doi:10.1093/eurpub/ckl060 Advance Access published on 3 May, 2006
............................................................................................................
Various
............................................................................................................
Disentangling gender and age effects on risky single occasion drinking during adolescence Emmanuel Kuntsche1, Gerhard Gmel1, Matthias Wicki2, Ju¨rgen Rehm3,4, Esther Grichting3 Objectives: To determine the association between adolescent risky single occasion drinking (RSOD) and gender, age, and the relative age position of students within their class. Method: A cross-sectional national representative sample of 7088 from 8th to 10th graders in Switzerland (mean age 14.6; SD ¼ 0.94) as part of the ESPAD international study was analysed using hierarchical linear modelling. Results: Being male, older than the class average and a member of older classes (on average) was associated with a higher RSOD frequency. Additionally, interactions between student and class level were found. The higher the mean class age the greater the difference in RSOD between boys and girls, and the lower the impact on RSOD of the relative age position in a class. Conclusions: In early and mid-adolescence, prevention efforts should try to impede the modelling of alcohol use of older classmates, whereas in late adolescence gender-specific motives for RSOD should be taken into account. Keywords: adolescents, old-for-grade, Switzerland, multilevel, group mean centring
............................................................................................................
Introduction Risky single occasion drinking (RSOD), also called ‘binge drinking’ or ‘heavy episodic drinking’,1,2 has increased among adolescents in many European countries.3 The increased prevalence of RSOD has raised public awareness of this problem and there have been calls for prevention strategies. However, RSO drinkers are not a homogeneous group of adolescents,4 and a ‘one size fits all’ intervention approach appears to be rather ineffective in reducing adolescent RSOD.5,6 It is therefore important to know which groups are prone to RSOD and how this behaviour develops over a life-time and particularly during adolescence. The present study focuses on school classes, an important environment in which to target prevention efforts. While it is common knowledge that gender and age are strong predictors of RSOD in adolescence and beyond,2,7,8 the present study investigates the development of the effects of gender and the relative age position of students in their class during their adolescent years by analysing the interaction of these effects with the mean age of the school class in hierarchical linear modelling (HLM). Evidence consistently shows that the prevalence of RSOD increases steadily during adolescence and peaks in late adolescence or early adulthood.8–13 Apart from the age effect on RSOD, it is claimed that the relative age position of students in their class has an effect on the level of RSOD. Studies from the US suggest that adolescents who are older than the majority of their classmates (called old-for-grade) are more likely than their classroom peers to use alcohol and other psychoactive substances.14,15 Byrd et al.,16 for example, revealed that old-
............................................................. 1 Swiss Institute for the Prevention of Alcohol and Drug Problems (SIPA), Research Department, Lausanne, Switzerland 2 Department of Psychology, University of Bern, Switzerland 3 Addiction Research Institute, Zurich, Switzerland 4 Public Health Sciences, University of Toronto, Toronto, Canada Correspondence: Emmanuel Kuntsche, Swiss Institute for the Prevention of Alcohol and Drug Problems (SIPA), Research Department, PO Box 870, CH 1001 Lausanne, Switzerland, tel: þ41 21 321 29 52; fax: þ41 21 321 29 40; e-mail:
[email protected]
for-grade adolescents are twice as likely to be risky single occasion drinkers. Accordingly, adolescents who are more physically mature than their same-age classmates are more likely to consume alcohol and other psychoactive substances.17,18 These older or more physically mature students may, in turn, serve as role models for the rest of the class as their alcohol consumption is often interpreted by younger students as a sign of independence and maturity.17,19 There are different reasons for being old-for-grade. Some students have their birthday near the beginning of the school term, others are retained by parents or teachers, sometimes owing to developmental problems. Academic failure that leads to repeating a school year is also a reason for being oldfor-grade. Byrd et al.,20 however, found that both grade-retained and not retained old-for-grade adolescents show higher rates of problem behaviour than their classmates. During adolescence, we hypothesize increasing gender differences and decreasing differences in RSOD as regards relative age. This means that, in early adolescence, students who are older or more mature than their classmates have a higher RSOD frequency but boys have the same frequency as girls. In late adolescence, both younger and older students in the class have the same frequency. However, boys have higher RSOD frequency than girls. Studies demonstrate that early physical maturation, for example, is likely to contribute to higher substance use rates particularly among girls.18,21 These differences in substance use are observed particularly in early adolescence, whereas they diminish in later adolescence.17 Furthermore, for different gender and age groups, distinct motivations to engage in RSOD were found. Enhancement motives, including drinking to get drunk or high, are particularly common among male drinkers.22–25 However, these gender differences do not develop until late adolescence. In her study, Cooper26 found no gender differences at the ages of 13 and 14, but a stronger increase in enhancement motives among boys in subsequent years. At the ages of 18 and 19, boys indicated a far higher level of enhancement motives than girls.26 Accordingly, in Switzerland, no gender differences in RSOD frequency were found among 11 and 13 year olds,27 whereas gender differences in RSOD are highly pronounced in the group of 15–24 year olds.13 Several longitudinal studies in the US found that girls were prominent in
European Journal of Public Health
groups with an early onset of RSOD, but the frequency of which did not increase during adolescence; boys were prominent in groups with an early onset of RSOD, the frequency of which increased over time.28–30 Therefore, we expect an interaction of gender and age in predicting RSOD frequency. We also expect the gap to close for relative age differences with increasing mean age within classes, because interindividual differences are expected to level off with increasing age owing to more normative patterns of RSOD. In the Adolescent Health Risk Study, a 5 year US follow-up study of 13–19 year olds, the increase in RSOD was steepest for the youngest cohorts and levelled off with age.12 No increase was found in the oldest cohorts (i.e. age from 19 to 24 and from 20 to 25 years). However, since alcohol consumption is legal at an earlier age in Switzerland (the legal drinking age is 16 for beer and wine, and 18 for spirits), we also expect age differences to level off at an earlier mean age than in the US.
Methods Study design and sample The database used for the analyses is part of the ‘European School Survey Project on Alcohol and Drugs’ (ESPAD).3 This survey has been conducted every 4 years since 1995 in about 30 European countries under the supervision of the Swedish Council for Information on Alcohol and Other Drugs and the Co-operation Group to Combat Drug Abuse and Illicit Trafficking in Drugs (Pompidou Group). Data were collected by means of a paper–pencil questionnaire which was administered in classes between the end of April and the end of June 2003. The time frame for filling out the questionnaires was one school lesson (45 min). Students could freely choose to participate and confidentiality was ensured at all stages of the study, e.g. by providing unmarked envelopes for completed questionnaires that could be sealed. Cluster sampling was used, based on a list of all classes of Swiss schools from 8th to 10th grade compiled by the Swiss Federal Statistical Office, where the classes served as the primary sampling unit. An overall response rate of 83.1% could be achieved (86.3% at class level and 95.9% at student level). Only 4.1% of the students in the participating classes did not take part in the survey because they were absent or refused. The 35 students (0.5%) who did not answer the RSOD question and the 32 students (0.4%) who did not indicate their age were excluded. The analysed data consists of 7088 13–17 year old students. This sample can be considered as representative for all 8th, 9th, and 10th graders in public schools in Switzerland. The total mean age was 14.6 years (SD ¼ 0.94).
Measures The questionnaire was developed by an interdisciplinary research group from the participating countries.3 Subsequently, the resulting questionnaire was translated under the supervision of SIPA in the three most frequently spoken languages in Switzerland: German, French, and Italian. RSOD. The adolescents were asked how frequently they had five or more drinks in a row in the past 30 days with the answers: ‘none’ (coded as 0), ‘1’ (coded as 1), ‘2’ (coded as 2), ‘3–5’ (coded as 4), ‘6–9’ (coded as 7.5), and ‘10 or more times’ (coded as 11, see e.g. Greenfield31 as an example to add one drink to the highest category). As the distribution on RSOD was skewed to the left, the logarithm was taken for the analyses in HLM.32 One drink was added before taking the logarithms [RSODLN ¼ LN(RSOD þ 1)], because the log of zero is not defined. Adding one drink puts the minimum useful value of the logarithmic transform back to zero.33 Age ranged from 13 to 17 years and was coded from 0 to 4, hence the intercepts corresponded to the values of the youngest
671
age group as a reference. Girls were coded 0 (boys ¼ 1) and thus girls made up the reference group.
Analytical strategy To estimate the impact of gender, age, and the relative age position of students in their class on RSOD, hierarchical linear models were estimated by means of the software package HLM 5.04.32 HLM also adjusts for design effects owing to clustering.34,35 To indicate the relative age of each student in a class, age was group mean centred, i.e. the deviation of each student from the mean class age. At the second level the mean age of each class was introduced.36,37 Cross-level interactions of gender and relative age with the mean age in each class at the second level were used to test the hypothesized increase in gender differences and the decrease in the relative age effects when adolescents grow older. To assess and compare the fit of the different models, Akaike’s Information Criterion (AIC) and Schwarz’s Bayesian Information Criterion (BIC) are applied.36 The first model includes only variables at the individual level. The formula is: RSODij ¼ b0j þ b1j ðgenderÞ þ b2j ðagegroup mean centred Þ þ r ij with b0j ¼ g00 þ u0j ðmean class ageÞ þ u0j and b1j ¼ g10 þ u1j and b2j ¼ g20 þ u2j where j indicates the class and i the individual within a class. The second model estimated the joint effect of individual level variables (i.e. gender and the relative age position in the class) and the class level variable (i.e. the mean age). The formula is: RSODij ¼ b0j þ b1j ðgenderÞ þ b2j ðagegroup mean centred Þ þ r ij with b0j ¼ g00 þ g01 ðmean class ageÞ þ u0j and bij ¼ g10 þ u1j and b2j ¼ g20 þ u2j where j indicates the class and i the individual within a class. The third model additionally included all cross-level interactions. The formula is: RSODij ¼ b0j þ b1j ðgenderÞ þ b2j ðagegroup mean centred Þ þ r ij with b0j ¼ g00 þ g01 ðmean class ageÞ þ u0j and b1j ¼ g10 þ g11 ðmean class ageÞ þ u1j and b2j ¼ g20 þ g21 ðmean class ageÞ þ u2j
Results A description of RSOD according to gender and age is given in table 1. More than one third of the total sample had already had an occasion in the past 30 days where five or more drinks were consumed in a row. Among boys the prevalence is about one
Table 1 RSOD (at least once in the last 30 days) according to gender and age—in percentage terms Total Boys Girls (N ¼ 7088) (n ¼ 3486) (n ¼ 3602) Total sample (N ¼ 7088)
35.9
42.1
29.8
.. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .
Age groups
.. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .
13 year olds (n ¼ 786)
20.9
20.1
21.6
14 year olds (n ¼ 2510) 31.9
36.9
27.1
15 year olds (n ¼ 2597) 40.7
49.4
32.4
16 year olds (n ¼ 1031) 43.5
51.7
35.2
17 year olds (n ¼ 164)
49.4
36.8
.. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. . .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. . .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. . .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .
42.7
672
Gender and age effects of RSOD
Table 2 Parameter estimates of the three hierarchical linear models estimated Individual level
Class level
Intercept (b0j)
Intercept (g00)
Individual level-only model
Individual and class level model
Cross-level interaction model
B
SE
t-ratio
B
SE
t-ratio
B
SE
t-ratio
0.348***
0.013
25.9
0.182***
0.027
6.9
0.258***
0.030
8.6
0.102***
0.015
6.7
0.054**
0.017
3.1
0.175***
0.019
9.2
0.001
0.039
0.0
0.108***
0.023
4.8
0.177***
0.033
5.3
0.061**
0.020
3.0
. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .
Mean age (g01)
. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .
Gender (b1j)
Intercept (g10)
0.173***
0.019
9.2
. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .
Mean age (g11)
. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .
Relative age (b2j)
Intercept (g20)
0.075***
0.013
5.7
0.075***
0.013
5.7
. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .
Mean age (g21)
. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. .. .. ... .. .. .. .. .. .. ... .
Model fit (AIC, BIC)
13816.2, 13864.2
13784.5, 13832.6
13779.0, 13827.1
**P < 0.01 ***P < 0.001
1.75 1.50
Frequency RSOD
1.25 1.00 0.75 0.50 female male total
0.25 0.00
13.0 to
